The goal of this proposal is to facilitate post-genomic era research by creating a protein structure prediction meta-server for close and distant homology modeling (HM).The proposed meta-server will be a fully automated, fast, widely available, autonomous, multi-component system, aimed at significantly increasing the number of proteins that can be characterized in-silico using current methods. Currently, successful distant HM requires considerable human expertise, including the manual detection and selection of templates and of sequence-template alignments using various fold-recognition (FR) methods, the adjustment of alignments and the refinement, evaluation and selection of alternative models. We aim at automating this process, using state-of-the-art FR meta-prediction technology. We propose 5 specific aims: 1) Evaluate the performance of various top-of-the-line, autonomous FR methods in order to identify the most accurate ones; 2) Develop improved algorithms to generate more accurate FR hybrid models and to better select FR- generated sequence-template alignments; 3) Develop improved methodologies to integrate the information from the FR models and/or the selected alignments to obtain more accurate refined full-atom models; 4) Test, optimize and implement all external software in-house; and 5) Apply the meta-server to specific problems of biological, experimental interest. From the users' perspective this meta-system will facilitate accomplishing one of the goals of Structural Genomics projects, namely, to better exploit the structural information from the experimentally solved proteins in order to obtain relatively accurate computational models for the majority of the remaining proteins, including those at close and those at distant HMdistances. The long term goal is to create a meta-server that will become a standard and will revolutionize the protein structure prediction field in an analogous way to the revolution that PSI-BLAST brought about in the sequence-comparison field, a few years ago.